Introduction
According to one common argument, the capacity of states to control immigration through migration policies is limited; after all, migration flows have gradually increased even as migration policies have become stricter (Castles 2004; Hollifield, Martin & Orrenius 2014). Other scholars argue quite the opposite – that nation-states have acquired more and better means of controlling their borders (Freeman 1994). Indeed, previous empirical analyses have shown that national migration policies have influenced immigration patterns in various contexts (Brekke, Roed & Schone 2016; Fitzgerald, Leblang & Teets 2014; Hatton 2004; Helbling & Leblang 2019; Ortega & Peri 2013).
Yet, many previous quantitative studies of the determinants of immigration have analysed bilateral migration flows between specific sending and receiving countries (e.g. Fitzgerald, Leblang & Teets 2014; Helbling & Leblang 2019; Mayda 2010; Ortega & Peri 2013). As such, these studies fail to capture the effects of policies or of other factors on the different entry categories (humanitarian, labour and family) that are the established categories in connection with international migration and which form the foundation on which national migration policies are constituted. A point of departure in this study is therefore that, if we are to account for the effects of different policies on immigration, we must examine different entry categories and devise different explanations for them (cf. Boräng 2018; de Haas 2021).
Against this background, this study sets out to analyse how admission policies and other factors influence patterns of family migration. Family migration has been the largest admission category in OECD countries during the past few decades, accounting for almost 40% of the total inflow in many countries (OECD 2019). It has also been increasingly contested in political debate in many countries, especially in the aftermath of the ‘refugee reception crisis’ of 2015 (Eggebø & Brekke 2019; Wray et al. 2023). Moreover, policies regulating family migration have become increasingly differentiated – as well as more restrictive overall – in many countries since the turn of the 21st century (Ahlén 2022; Helbling et al. 2017). Yet, notwithstanding the significance of family migration as a regular type of entry, little scholarly attention has been paid to it compared to other categories of entry, such as labour and humanitarian migration (Bonjour & Kraler 2015). Not only have very few studies analysed how policies influence different categories of entry; no comprehensive cross-country comparisons have so far been done of the impact of different policies on patterns of family migration.
This paper makes a first attempt to address this gap by analysing the connection between variations in admission policies and the inflow of family migrants in 29 to 31 European states during the 2008 to 2019 period. Newly disseminated data from Eurostat (2021a) and the Migrant Integration Policy Index (MIPEX) (Solano & Huddleston 2020) make it possible, on the one hand, to separate different external and internal policies on family migration from each other and, on the other, to distinguish between different sub-groups among family migrants. My aim here is to investigate whether and how admission policies and other factors have influenced different types of family related immigration. More specifically, I am interested in whether admission policies have resulted in stratifying outcomes, whereby certain groups have been affected more than others. In order to examine this question, I differentiate between two kinds of family migration. In one case, the resident sponsor is either an EU citizen (i.e. a national of the destination country or a mobile EU citizen); in the other, he/she is a non-EU citizen (i.e. a third-country national [TCN]). In both cases, however, the incoming family member is a TCN (i.e. a non-EU citizen). As discussed further below, we can assume these two groups of resident sponsors enjoy differing degrees of inclusion in the receiving country. Accordingly, the degree of restrictiveness in admission policies can be expected to affect these groups in different ways. Furthermore, as elaborated below, there exists some inconsistency between national and EU laws concerning the right to family (re)unification. This may complicate the categorization of resident sponsors into these two sub-groups. However, I contend that despite this complexity, it remains a practical way to identify groups likely to face varying challenges when policies become stricter.
What Factors Influence Family Migration Rates?
Admission Policies
Family migration, as defined here, includes immigration by TCNs who have obtained residence for family reasons in a European country. The category of family migration includes two main sub-categories: family reunification and family formation. ‘Accompanied family members’, a third type of family migration, refers to the process whereby the members of a family migrate together – that is, where a migrant is accompanied by his/her ‘dependents’ (Kofman 2004). Since other rules and requirements usually apply for this type of movement (as in the case of seasonal workers, for instance), it is not included in the analysis here.
Family reunification refers to the process whereby a principal migrant who has settled in a receiving country reunites with family members living abroad. The second category, family formation (or so-called ‘marriage migration’), occurs when an individual in the receiving country chooses a partner from abroad (Kofman 2004). While family formation makes up an increasing part of family migration in OECD countries, family reunification is the most common type (see, e.g. OECD 2017: 115ff). Since specific permit data that distinguishes between cases of family reunification and of family formation are only available in a few countries (see OECD 2017), it is not possible to conduct any broader cross-country analysis of differences between these two forms of family migration.
National policies regulating the admission of family immigrants who are TCNs include two sub-dimensions: eligibility criteria and qualifying conditions. Eligibility criteria regulate who can apply for family migration. On the one hand, they concern the legal status of sponsors. Must they have a certain type of residence permit before their family members can immigrate? Must they have resided in the country for a certain length of time? On the other hand, eligibility criteria regulate which types of family members are allowed to unite with their sponsors, for example, whether entry is granted to relatives other than spouses (Bjerre et al. 2016).1 Qualifying conditions are additional requirements that eligible applicants must fulfil in order to gain admission. Sponsors, for example, may need to prove they possess sufficient financial resources and that they are able to furnish adequate accommodations for the new arrival(s), while incoming family members may need to demonstrate minimum language skills and some knowledge of the country (Bjerre et al. 2016).
Since family migration concerns a resident sponsor who seeks to (re)unite with one or more family members living abroad, different criteria usually apply depending on the legal status of said sponsor, for example, whether he/she is a citizen or has acquired a residence permit (Bonjour & Kraler 2015). If the resident sponsor is an immigrant, the prospects for family (re)unification can vary according to the type of residence permit obtained, for example, whether the sponsor has temporary or permanent residence, or whether he/she has gained admission to the country based on work or for humanitarian reasons. Thus, various rights and regulations go together with a particular legal status on the part of the sponsor, prompting some authors to speak of a ‘politics of belonging’ or a ‘hierarchy of stratified rights’ (Block 2015; Kraler et al. 2011). Moreover, the insider/outsider overlap facilitates the use of additional policies to control admission. Most importantly, requirements can be placed both on resident sponsors and on incoming family members. A key aspect of this double conditionality is that demands for integration, such as income requirements aimed at resident sponsors, can be used to pursue goals of immigration control (Bech, Borevi & Mouritsen 2017).
Given the multi-dimensionality of admission policies and the variations in recent policy reforms, we can expect admission policies to influence inflows of family immigrants. Yet, considering that many have called the capacity of states to control immigration into question, this expectation is not self-evident. The effects of various policies may also be futile or perverse. The first aim of my analysis here, therefore, is to investigate whether – and if so, how much – restrictive admission policies have led to lower inflows of family immigrants (H1). Accordingly, I expect both eligibility criteria and qualifying conditions to influence inflows.
Stratifying Implications of Admission Policies
One typical argument in the literature is that family migration policies for TCNs have become not just more restrictive across countries, but increasingly conditional as well, such that many European countries have introduced demanding admission requirements for family migration (Bech, Borevi & Mouritsen 2017; Bonjour & Kraler 2015). This has spurred debates about stratified outcomes, whereby policies have different effects on different types of family related admission (e.g. Block 2015; Goodman 2011; Scholten et al. 2012; Schweitzer 2015; Sirriyeh 2015; Strik, de Hart & Nissen 2013; Wray 2009). Goodman, for example, has argued that pre-arrival civic-integration policies have dis-incentivized family based migration and led to a selection of migrants that are easier to integrate (2011; 2014). In a similar vein, Kofman (2018) argues that regulating family based admission based on income may have the effect of choosing the ‘best and brightest’ while excluding those on low incomes and with few resources.
However, while many have highlighted the potential stratifying effects of recent policy reforms, there is very little evidence on the actual influence of policies on admission. Scholten et al. (2012) have shown that the introduction of language tests has served to raise the educational level of spouses admitted to Germany and the Netherlands. Even so, no comprehensive cross-country comparisons have been done of how policies and other factors have shaped the size and composition of family migration over time.
One way to address this gap is to use new data from MIPEX (Solano & Huddleston 2020) and from Eurostat (2021a). This data allows us to differentiate between the admission of TCN family members according to whether their resident sponsors are EU citizens (including nationals) or non-EU citizens. This means that, for all first permits granted to TCNs for family reasons, it is possible to differentiate between, on the one hand, resident sponsors who are citizens of the destination country or another EU country and, on the other hand, resident sponsors who are not citizens of the destination country or another EU country (i.e. who are immigrants from outside Europe).
For additional clarity, the definition and measurements of family migration applied in this study only include TCNs who have been granted residence permits for family reasons in a European country. Resident sponsors, however, can be of any nationality. The distinction that applies is whether the sponsor is an EU citizen or a non-EU citizen – that is, a TCN with a valid residence permit in a receiving European country. Thus, family migration involving only EU citizens (i.e. internal mobility within the EU) – as for instance, when a German citizen moves to Sweden to reunite with a German citizen already living in Sweden – is not included. Table 1 delineates the family migration categories and the combinations of resident sponsors and incoming family members considered in this study.
Table 1
Summary of categories and combinations of family migration included in this study.
| INCLUDED | EXCLUDED | NOTES | |
|---|---|---|---|
| Sub-categories of family migration | Family reunification Family formation | Accompanying family members | The testing includes both family reunification and family formation cases but does not make a distinction between them. |
| Resident sponsors | EU citizens (incl. nationals) Third-country nationals | The testing includes and distinguishes between resident sponsors who are EU citizens (i.e. nationals of the destination country or mobile EU citizens) and non-EU citizens (i.e. third-country nationals that have a residence permit in the destination country). | |
| Incoming family members | Third-country nationals (non-EU citizens) | EU citizens | The testing only includes incoming family members that are third-country nationals (i.e. from outside the EU). |
Since the available data only allows us to distinguish between these broad categories of resident sponsors (EU citizens vs. non-EU citizens), it is necessary to highlight and problematize what Staver (2013) calls the ‘fragmentation of family reunification rights’. As Staver (2013) stresses, there is a degree of mismatching between national and EU legislation when it comes to the right to family reunification. This can cause ‘reverse discrimination’, whereby citizens of a country with strict family migration policies are disadvantaged with regard to family reunification as compared with mobile Europeans living in the same country, who instead are treated in accordance with more generous EU legislation (Staver 2013: 70). Although the focus here is on differences between EU sponsors and non-EU sponsors, the mismatch can mean that different rules apply to different sub-groups among EU sponsors (e.g. to nationals and mobile EU citizens) in some countries, which may then weaken the reliability of the measurements applied. However, the group of mobile EU citizens reuniting with TCNs is comparatively small. In 2018, for example, 1,977 TCN family members were united with mobile EU citizens residing in Sweden; this represents approximately 3% of the total number of residence permits issued to TCNs for family reasons that year (68,068 cases in total) (Swedish Migration Agency 2021). Thus, the distinction between EU and non-EU sponsors still serves as a feasible proxy for differentiating between groups that on a collective level can be expected to face differing difficulty when policies become more demanding.
From the standpoint of admission, this distinction between sponsors according to their citizenship status can be of vast importance for their prospects of uniting with family members from abroad. Previous studies have claimed that, due to the expanded elements of stratification in family migration policies, nationality now increasingly determines – in combination with socioeconomic and sociocultural factors – whether it is more or less difficult (or at times even impossible) for individuals to reunite with their family (Bonjour & Duyvendak 2018; Kofman 2018). As Block (2015) emphasises, the idea of a right to family migration is permeated by notions of who belongs (more) to the national community, which creates hierarchies of membership among claimants to family unification. Thus, ‘the stronger your membership and the “better” a member you are, the more rights you can claim and access’ (Block 2015: 1439).
From this perspective, eligibility criteria and qualifying conditions can be seen as demanding membership requirements that both directly and indirectly affect certain sub-categories among family immigrants more than others. Where residence requirements for resident sponsors (eligibility criteria) are concerned, we may speak of formal stratification. It is naturally harder for sponsors who hold temporary residence permits to comply with stricter requirements regarding the length of time they need to have lived in the country before family reunification is allowed. Such a status is likely in the case of many non-EU sponsors who have themselves immigrated from outside Europe. But it does not apply, of course, to the citizens of the country in question.
Where qualifying conditions are concerned, we can likewise expect strict income requirements to be more difficult for non-EU sponsors who have themselves immigrated from outside Europe than they are for EU nationals (including nationals of the country itself), who are likely to enjoy a higher degree of inclusion (Block 2015; Wray 2009). This relates to what Morris (2003: 87) has called ‘informal stratification’, whereby demanding conditions are harder to meet for applicants with less in the way of financial, social, or cultural capital (see also Schweitzer 2015).
Accordingly, restrictive admission policies are likely not only to reduce the overall inflow of family migrants, but also to have a stratifying impact among them. Given the characteristics of the different sub-groups among family migrants, I hypothesise that restrictive admission policies will be more effective in regulating family migrant inflows when the sponsors are TCNs than when they are EU citizens (H2).
Other Factors
Apart from being regulated through external admission policies (eligibility criteria and qualifying conditions), family migration is also regulated internally – through policies that stipulate how long family immigrants can stay in the country and how secure their status is (Bjerre et al. 2016). Variations in the type of residence permit granted to family immigrants – regarding duration, possibilities for renewal, grounds for acceptance or rejection, and so on – determine the security of the status enjoyed by these persons. Variations of this kind can play an important role in the decision made by such persons to apply for family reunification in a particular receiving country.2
Besides policies, other features of a destination country may be important in shaping as well as in attracting or deterring family migration. Previous studies have shown that various demographic, economic and political factors influence migrant inflows (Brekke, Roed & Schone 2016; Fitzgerald, Leblang & Teets 2014; Hatton 2004; Helbling & Leblang 2019; Mayda 2010; Ortega & Peri 2013). While these studies have focused on inflows of irregular migrants or on bilateral migration between specific sending and receiving countries, some of the factors discussed below may also influence inflows. Moreover, these factors may not only shape patterns of family migration; they may also confound the effect of admission policies on inflows.
It is widely acknowledged in migration research that earlier immigration from the same origin paves the way for continued migration to a particular destination (de Haas 2010; Massey et al. 1993). Family migration, moreover, adds another dimension to ‘social-network theory’. Since previous immigration is a prerequisite for family reunification (which is the most common type of family migration), a large stock of foreign-born persons may be expected to function as a particularly strong pull factor for family migration. In contrast, family reunification is less likely to occur in countries with a low number of foreign-born persons.
In functionalist neoclassical migration theory, according to which migrants mainly seek to maximise utility, the economic wealth of a country is regarded as the main factor determining international migration (Fitzgerald, Leblang & Teets 2014; Massey et al. 1993; Ortega & Peri 2013). Economic drivers of migration are most often associated with the mobility of international workers, but they can also influence other forms of cross-border movement. In the case of family migration, a high level of economic wealth may be expected to attract both family formation and family reunification to a particular location. In contrast, a high rate of unemployment is likely to deter immigration, because it indicates the overall economic situation is poor and that job opportunities are scarce (Helbling & Leblang 2019).
A similar logic of attractiveness may be connected with the welfare state. According to the welfare-magnet hypothesis, one of the main determinants of migrants’ choice of destination is the relative generosity of the receiving state’s social benefits (Borjas 1989). Hence, the generosity of the social-welfare system – represented in this study by the level of social assistance – can be expected to constitute a pull factor for family migration.
Where the political environment is concerned, previous studies have argued that the rise of anti-immigration parties has not only increased the pressure to enact restrictive policy reforms; it has also affected the attractiveness of potential destinations (Fitzgerald, Leblang & Teets 2014; Gudbrandsen 2010). Fitzgerald, Leblang and Teets (2014) find that the strength of radical-right parties influences migration patterns: ‘A country with electorally competitive radical right parties will receive smaller flows of migrants because this political characteristic signals that the social and political fabric is not supportive of incorporation’ (p. 407). Thus, apart from being a potential driver of restrictive policy changes, an anti-immigration political environment – which may be represented by the electoral success of political parties with explicit anti-immigration agendas – has a deterring effect on the inflow of family migrants.
While these demographic, economic and political factors can be expected to influence family migration flows, they may also shape a country’s migration policies. Previous research has shown that, depending on their institutional design, different welfare states have been more inclined or less to introduce stricter policies for family migration when under pressure from increasing immigration and higher unemployment (Ahlén 2023). My concern with these connections, therefore, is with the potential confounding influence of these factors on the relationship between admission policies and family migration. To address this concern, as described below, I combine fixed-effects regression models with various robustness tests, in order to isolate the relationship between admission policies and inflows of family immigrants.
Research Design
Data
Data on migration policy is taken from the MIPEX, which released updated data in December 2020 for 52 countries between 2007 and 2019.3 While MIPEX primarily provides data on integration-related policies, it also contains indicators of policies that regulate family migration.4 Here, it is important to note that these MIPEX indicators measure the regulation of family reunification when the resident sponsor is a TCN; as such, they do not take formal account of the regulation of family migration when the sponsor is a citizen of the destination country (see Solano & Huddleston 2020). However, since the Court of Justice of the European Union has ruled that EU member states may not make the right of family (re)unification dependent on the citizenship of the sponsor (see, e.g. COM (2019) 162 final),5 it can be assumed that equal conditions generally apply for family migration where the sponsor is either a national citizen or a TCN. On the other hand, as previously noted, some member states have slightly different requirements for TCN sponsors than for citizen sponsors (see, e.g. European Migration Network 2017; Staver 2013). This is important to keep in mind when interpreting the results, especially regarding the effects of policies on different sub-groups among family immigrants.
All of the policy variables measure the degree of inclusivity/restrictiveness in a set of indicators on a scale of 0 to 100, where higher values represent more inclusive and less restrictive policies, while lower scores indicate more restrictive policies (the scale of policy restrictiveness applied in the test is reversed relative to that in the original scale in MIPEX). The scoring methodology involves a set of indicators within each policy area. These indicators are evaluated against a set of standards or benchmarks based on international norms or best practices. The scores are then calculated based on the extent to which a country’s policies align with these benchmarks (Solano & Huddleston 2020).
Policies that specifically target family based immigration by TCNs are distinguished along two dimensions. Admission policies (external regulation), which are the main independent variables, include the two sub-dimensions of eligibility criteria and qualifying conditions, each of which is aimed at either resident sponsors or incoming family members. Residence-permit policies (internal regulation) regulate the type of residence permit granted to a given type of family immigrant in connection with duration, possibilities for renewal and grounds for acceptance or rejection.6
Data on immigration rates and stocks of foreign-born persons covering 31 European countries for the 2008 to 2019 period is taken from the Eurostat database (Eurostat 2021a). Rates and stocks both concern TCNs and are estimated per 1,000 population. Three different measures of inflows are applied: Total inflow of family migrants; Inflow of family migrants where the resident sponsor is an EU citizen (including nationals of the receiving country); and Family immigrants where the sponsor is a TCN (non-EU citizen).
Immigrant inflow other than family immigrant inflow is used in two ways. Total Immigrant inflow (excluding family immigrant inflow) is used to control for all other push and pull factors that can influence family migration, as well as other types of immigration to a country. This measurement is also applied as a placebo outcome in a falsification test (Model 14). Stocks of foreign-born are included as a control variable in the regression analysis.
GDP per capita (Eurostat 2021b) is used to capture differences in wealth and in overall economic performance. The unemployment rate is estimated as the share (percentage) of unemployed persons in the labour force between the ages of 15 and 64 (ILO 2020). Information on Social-assistance rates is taken from the newly released update of the Social Assistance and Minimum Income Protection Interim Dataset (SAMIP) (Nelson et al. 2020). The combined measurement ‘SAavey’ estimates social-assistance standard rates in yearly amounts for (1) single adult persons below retirement age and without children, (2) single parents with two children and (3) two-parent families with two children.7
Two variables in the Varieties of democracy dataset (V-dem) (Coppedge et al. 2021) are combined in order to estimate the Vote share of anti-immigration parties: ‘v2paimmig_mean’ (‘What is the party’s position regarding immigration into the country?’) and ‘v2pavote’ (‘Vote share the party gained in the election to the lower chamber’). Any party that scores lower than 1 on ‘v2paimmig_mean’ (‘this party strongly opposes all or almost all forms of immigration into the country’) is categorised as an anti-immigration party. The measure used in this test is the percentage of the vote won by these parties in the most recent national election (‘v2pavote’).
In the following analyses, I use an index that combines these data sources on policy, migration, and the other variables for 31 countries for the 2008 to 2019 period.8 The focus on European countries is motivated by the regulatory definition of immigration from a third country, and it follows the delimited data on immigration stock and rates provided by Eurostat (2021a). Descriptive statistics for all variables are presented in Table A1 in the Appendix.
Method
Using panel data, I perform a time-series regression analysis with fixed effects (FE) and standard errors clustered at the country level in order to assess changes in family migration in 29 to 31 European countries during the 2008 to 2019 period (Table 2). Given the diversity of drivers and opportunities that shape migration processes in both macro and micro terms, any study of the effects of different policies on inflows needs to take the possibility of endogeneity (i.e. when an explanatory variable is correlated with the error term) seriously. Aware of this challenge, all models include country- and year-FE, including a variety of theoretically selected control variables, in order to test how the predictor variables influence within-country changes in family immigrant inflows during the period studied. Although the period is rather short (2008–2019), I make use of this strategy in order to control for time-invariant country-level factors and to avoid omitted-variable bias (Allison 2009).
Table 2
Fixed-effects regression models of family migration rates in 29–31 European countries, 2008–2019.
| FAMILY MIGRATION (ALL) | FAMILY MIGRATION (ALL) | FAMILY MIGRATION (ALL) | SPONSOR EU CITIZEN | SPONSOR NON-EU CITIZEN | |
|---|---|---|---|---|---|
| MODEL 1 | MODEL 2 | MODEL 3 | MODEL 4 | MODEL 5 | |
| Admission policies t – 1 | –0.0173** (0.00688) | –0.0198** (0.00936) | –0.0192** (0.00889) | –0.00257 (0.00354) | –0.0149** (0.00678) |
| Residence-permit policies t – 1 | 0.00417 (0.00608) | 0.00388 (0.00641) | –0.00312 (0.00290) | 0.00494 (0.00489) | |
| Social-assistance rates t – 1 | 0.00006* (0.00003) | –0.00002 (0.00002) | 0.00006 (0.00004) | ||
| Stocks of foreign-born t – 1 | 5.674* (3.324) | 1.989** (0.933) | 6.235** (2.677) | ||
| Immigrant inflow (excl. family) | 0.0399** (0.0185) | –0.0070** (0.0034) | 0.0401** (0.0174) | ||
| Unemployment rate t – 1 | –0.0245 (0.0334) | –0.0189 (0.0152) | –0.00628 (0.0358) | ||
| Vote share of anti-immigration parties | –0.00257 (0.00355) | –0.00095 (0.00184) | –0.00005 (0.00384) | ||
| GDP per capita (log) t – 1 | –0.469 (0.599) | –0.108 (0.307) | –0.0167 (0.543) | ||
| Constant | 1.968*** | 1.833*** | 5.315 | 2.004 | –0.121 |
| (0.254) | (0.252) | (6.028) | (3.109) | (5.446) | |
| Mean of dependent variable | 1.61 | 1.61 | 1.61 | 0.56 | 1.05 |
| Observations | 320 | 320 | 288 | 283 | 283 |
| R-squared | 0.252 | 0.253 | 0.346 | 0.213 | 0.324 |
| Number of countries | 31 | 31 | 29 | 29 | 29 |
[i] Notes: Fixed-effects regression models with clustered standard errors. Immigration rates and stocks of foreign-born concern third-country nationals (TCNs) and are measured per 1000 population. The policy variables are measured from 0–100, where higher values symbolise a higher degree of restrictiveness and lower values a lower one. The reason for not lagging Vote share of anti-immigration parties was to avoid the risk of incorrectly associating changes in family migration with a former political environment and previous election results. ***p < 0.01; **p < 0.05; *p < 0.1.
Model 1 tests the association between admission policies and family migration rates alone. Model 2 includes residence-permit policies as well. Model 3 tests the additive effects of all independent variables on family migration rates. The same model specification is used for assessing the aggregated measurement of family migration rates, as well as of family migration where the sponsor is an EU citizen (Model 4) and where he/she is a non-EU citizen (Model 5), separately. The resulting model can be specified as follows:
Here, Yit represents the rate of family migrants in country i at year t. β1Ait – 1 is admission policies and β2Rit – 1 is residence-permit policies. β3Sit – 1 is social-assistance rates; β4Fit – 1 is stocks of foreign-born; β5Iit is rates of immigrants other than family immigrants; β6Uit – 1 represents unemployment rates; β7Vit is the vote share of anti-immigration parties; and β8Git – 1 is GDP per capita. σi represents the inclusion of country-FE and θt the inclusion of year-FE. ∈it is the error term.
Robustness tests of the relationship between admission policies and family migration rates are carried out in Models 6 to 9 in Table A2 and in Models 10 to 15 in Table A3 in the Appendix. Models 6 and 7 assess the effect of admission policies on family migration rates in two different country groups: Northern and Western European countries9 (Model 6), and Central/Eastern and Southern European countries10 (Model 7). Model 8 tests the effect of eligibility criteria on family migration rates, while Model 9 tests the effect of qualifying conditions.
Models 10 to 15 in Table A3 assess the robustness of the relationship between admission policies and family migration rates. Models 10 to 12 mirror Models 3 to 5 in Table 2; but instead of using FE, they include lagged dependent variables (LDVs) in order to control for past values of the family migration rate. Given the potential for negative weight issues in two-way FE estimation, especially when average treatment effects (ATEs) vary across distinct groups or periods (de Chaisemartin & d’Haultfoeuille 2020), incorporating LDVs is a helpful strategy. LDVs capture the past influence of the dependent variable on the current outcome, effectively controlling for unobserved factors that could impact both current and lagged values (Donald and Lang 2007). Moreover, when dealing with a continuous treatment measure instead of a binary one, integrating LDV models helps addressing the complexities of temporal dynamics (Angrist & Pischke 2009). This approach assists in mitigating interpretation challenges associated with continuous treatment measures (see Callaway, Goodman-Bacon & Sant’Anna 2021).
However, as Nickell (1981) first noted, including both FEs and LDVs has the drawback of producing biased parameter estimates. Following Angrist and Pischke (2009, p. 245), therefore, I use FE and LDVs as alternative identifying assumptions in separate models in order to bound the causal effect of admission policies on family migration rates and at the same time to avoid a ‘Nickell bias’.
In Model 13, the variable of admission policies is lagged 2 years (t – 2), in order to account for the delayed effects of policy changes on inflow patterns. Model 14 applies inflows of immigrants other than family migrants, as a placebo outcome, in order to check the validity of the relationship between admission policies and family migration. In Model 15, admission policy is treated as the dependent variable and family migration as the independent variable, in order to assess concerns about reversed causality. The model also includes the other independent variables in order to control for their potential confounding influence.
All in all, by combining fixed-effects models with various robustness tests based on different identification assumptions (cf. Keele 2015), this analysis controls for spurious correlations and potential confounders. This makes it possible to subject the causal arguments in question – regarding the relationship between admission policies and family migration rates – to a proper test.
Results
Table 2 presents the estimations of the time-series regression analysis of family migration rates in 29 to 31 European countries during the 2008 to 2019 period.
The Influence of Admission Policies on Family Migration Rates
Models 1 to 3 show that more restrictive admission policies are associated with falling family migration rates. The admission-policies coefficients range between –0.017 and –0.02, which in effect means that an increase by 1 point in policy restrictiveness (scale 0–100) generates a fall in family migration rates by 0.0173 to 0.0198 persons per 1,000 population. This can be compared with the mean value of 1.61 residence permits granted to TCNs for family reasons per 1,000 population in 31 European countries between 2008 and 2019.
To illustrate what this result means, we can consider the case of the UK. With similar a mean value of the dependent variable as the cross-country average (1.61 permits granted to TCNs for family reasons per 1,000 population), we can estimate how a policy reform affects the inflow of family migrants according to the results presented in Table 2. In 2011, the UK introduced a pre-entry language test for family members abroad (Kofman 2018). This policy shift (which concerns one out of six indicators making up the variable of admission policies) generated a positive change in the overall score by 6.6 points. Assuming the coefficient of admission policies in Model 3 (–0.0192), this reform would have the effect of reducing the number of family immigrants to the UK (with a population of roughly 63 million in 2011) by approximately 8,000 persons in the year after the new policy was implemented.
As Model 3 shows, the influence of admission policies on family migration is statistically significant when the other predictor factors are controlled for. This result corroborates H1, that is, the hypothesis that restrictive policy changes will have a negative effect on family immigrant inflows. Moreover, different geographical patterns can be distinguished here. Models 6 and 7 in Table A2 in the Appendix reveal inverse tendencies between two country groups. While a statistically significant negative relationship between admission policies and family migration rates is found in Northern and Western European countries, the relationship is positive (albeit not significant) in Central/Eastern and Southern Europe. This reflects the findings of previous studies, according to which policies have a real impact in destination countries characterised by contextual conditions that are likely to attract further immigration (e.g. Helbling & Leblang 2019).
Where the two sub-dimensions of admission policies are concerned, both eligibility criteria and qualifying conditions are negatively correlated with family immigrant inflows. The effect of eligibility criteria is much larger, however, and it is statistically significant as well, which is not the case with qualifying conditions (see Models 8 and 9 in Table A2 in the Appendix).
Taken all together, the models in Table 2 support the claim that admission policies influence family immigrant inflows. This conclusion is supported, moreover, by the result of the robustness tests in Models 10 to 15 (Table A3). The coefficients of the LDVs in Models 10 to 12 reveal that family migration rates are quite strongly associated with their past levels, particularly for the total inflow of family migrants (Model 10) and for family immigrants joining a non-EU citizen (Model 12). Still, this also furnishes support for the other predictor variables that have a significant influence on the outcome, such as admission policies in Models 10 and 12 (Keele & Kelly 2006). Here it should be noted that the coefficient sizes of admission policies are smaller in the models with LDVs (Models 10–12) than in the fixed-effects models (Models 3–5). This is because these estimates represent an underestimation and an overestimation of the true effects. Thus, according to Angrist and Pischke (2009), a more precise interpretation of the effect sizes lies between the estimates of these different models.11
Additionally, the LDV Models (10–12) help addressing the potential issue of negative weights in two-way FE estimations and the complexities of temporal dynamics and the potential persistence of treatment effects (cf. Callaway, Goodman-Bacon & Sant’Anna 2021; de Chaisemartin & d’Haultfoeuille 2020). Model 13 furthermore shows that admission policies (t – 2) have a delayed effect on family immigrant inflows. The result in Model 14 shows that family migration policies are not significantly associated with changes in immigrant inflows other than in the case of family migration. Model 15 further indicates there is no reversed causality between admission-policy restrictions and family migration rates.
It should be acknowledged that the R-squared values in all regression models are notably low. This underscores the empirical challenges involved in the kind of comparative analysis undertaken in this study. Migration is indeed an inherently unpredictable phenomenon, among other things, due to changes in people’s aspirations and opportunities, as well as in the ever-changing push factors. Moreover, establishing causal links in a comparative setting using observational data is challenging. While a higher R-squared generally suggests a better fit, a low R-squared does not necessarily invalidate the model. However, despite these challenges, the theoretically founded assumptions and statistically significant findings provide a good basis for drawing general conclusions about the effects of family migration policies. The different fixed-effects models and lag structures further help isolate the relationship between admission policies and family migration, thereby rendering the causal interpretation of the findings credible (Angrist & Pischke 2009; Keele 2015).
Coming back to the result of Model 14, there is another noteworthy observation concerning the effects of policies that deserves attention. While the falsification tests buttress the causal effect of interest, the result in Model 14 is also somewhat surprising, given that a plausible expectation would be that a country’s migration policy is relatively coherent – that is, that it is characterised by a similar degree of restrictiveness in relation to different types of immigration. On the other hand, this finding highlights the diversification of both migration policies and types of immigration. Restrictive reforms in one area of migration policy do not necessarily entail restrictions in other areas, and the effects of different policies can vary in relation to different categories and sub-categories of migration. This accords with the approach taken in this paper, namely, that we must differentiate between different policy tools and dimensions in connection with different immigration categories if we are to achieve analytical precision in assessing the relationship between policy outputs and outcomes (cf. Boräng 2018; de Haas 2021).
Stratifying Effects
Let us now turn to the stratifying effects. The differences between Models 4 and 5 in Table 2 give support to H2. Both the coefficient sizes and the significance levels show that restrictive admission policies have a stronger impact on family immigrant inflows when the sponsor is a non-EU citizen than when he/she is an EU citizen.
Thus, the analysis here shows that increasing policy restrictions not only affect overall inflows of family immigrants; they also lead to stratified patterns, whereby certain sub-categories among family immigrants are more heavily affected. This confirms the claim that many scholars have made regarding growing conditionality and stratification in family migration policy (e.g. Bech, Borevi & Mouritsen 2017; Block 2015; Bonjour & Kraler 2015; Goodman 2011; Kofman 2018; Scholten et al. 2012; Schweitzer 2015; Sirriyeh 2015; Strik, de Hart & Nissen 2013; Wray 2009). Moreover, by expanding the geographical and temporal scope of the investigation, my analysis here contributes new general insights on this issue.
While it can be assumed, as discussed above, that requirements are generally equal for family (re)unification whether the sponsor is a national citizen or a TCN, the available data does not make it possible to measure the impact of differing policies for different sub-categories among immigrants. Thus, these differential policy effects should be interpreted with some caution. Moreover, the available data on migrant inflows only distinguishes between categories among family immigrants according to sponsors’ citizenship. Using the difference between EU and non-EU sponsors as the only stratifying dimension is a limitation, but I contend that it still serves as a feasible proxy for differentiating between sub-groups that can be expected, on a collective level, to face differing degrees of difficulty when policies become more demanding. For example, stricter residential requirements for resident sponsors explicitly target sponsors who hold temporary residence permits, which is likely in the case of non-EU sponsors who have themselves immigrated from outside Europe. Nationals and EU citizens can also generally be expected to find it easier than TCNs to cope with strict income and housing requirements. TCNs, many of whom are newly arrived immigrants, have usually not reached the same degree of inclusion in the destination country (Strik, de Hart & Nissen 2013). We can accordingly expect the impact of restrictive admission policies to be greater in the case of family immigrants with non-EU sponsors.
Consequently, by introducing demanding admission policies that require applicants to have achieved a certain degree of inclusion, states can respond to public concerns over family reunification with TCN sponsors, while at the same time protecting the right of citizens to (re)unify with partners residing abroad. Such policies may also have a ‘dual functionality’ – controlling immigration and facilitating ‘better’ integration at the same time (Goodman 2014). This speaks to the increasingly negative framing of family migration in the political debate – especially the family reunification of refugees – as constituting ‘unwanted chain-migration’ that poses a threat to social cohesion and the welfare state (Bonjour & Duyvendak 2018; Kofman 2018).
Furthermore, while it is beyond the scope of this paper to analyse the political intentions underpinning certain policy reforms, the evidence furnished here offers additional insights into debates concerning how these stratifying effects are, or can be, justified by policy-makers in European democracies (Bech, Borevi & Mouritsen 2017; Bonjour & Kraler 2015; Wray et al. 2023).
Other Factors
Lastly, a few comments on the influence of the other predictor factors are warranted. The connection between residence-permit policies and family migration rates is inconclusive. While the relationship is negative in Model 4 (EU sponsors), it is positive in the other models. None of the coefficients, however, are statistically significant. Thus, the tests performed here do not support the idea of any direct connection between internal policies and family migration.
As expected, stocks of foreign-born and immigrant inflows of other kinds are associated with rising family migration rates (except in the case of other immigrant inflows in Model 4). While the positive influence of a large foreign-born population is in line with the ‘social-network theory’ (Massey et al. 1993), the positive relationship between family migration and other types of inflow highlights the significance of other underlying push and pull factors that shape migration flows to and from a particular country. Yet, by controlling for other types of inflow, the significance of the impact of admission policies on family migration rates seen in Models 3 and 5 is strengthened.
The negative associations between family migration rates and both higher unemployment rates and higher vote shares for anti-immigration parties are in line with my theoretical expectations. However, these relationships are not statistically significant. Higher social-assistance rates are associated with rising family migration rates (except when the sponsor is an EU citizen). In line with the welfare-magnet hypothesis, this finding supports the view that a generous social-welfare system is an attracting factor for family migration. In contrast (and surprisingly), a higher GDP per capita is associated with falling family migration rates.
Concluding Remarks
This study has contributed new insights on whether and how much migration policies influence immigration. The empirical evidence set forth here concerns the connection between admission policies and family immigrant inflows in 31 European countries between 2008 and 2019. Two hypotheses on the effects of admission policies on family immigrant inflows have been presented.
The results reached herein show that changes in admission policies have shaped the size and composition of family immigrant inflows. While restrictive admission policies are associated with falling family migration rates overall, my analysis reveals that admission policies have had a stronger impact on family migrant inflows when the sponsor is a non-EU citizen than when he/she is an EU citizen. As the first comprehensive cross-country examination of the effects of family migration policies, this study expands the geographical and temporal scope of analysis beyond that found in previous research in this area, which to date has been dominated by in-depth case studies of a small number of countries (e.g. Bech, Borevi & Mouritsen 2017; Sirriyeh 2015; Strik, de Hart & Nissen 2013). While additional case studies could be beneficial in delineating policy effects across different settings, the findings presented herein contribute new general knowledge on the differential impact of admission policies on patterns of family migration over time.
Whether policy-makers have intended it or not, stricter eligibility criteria and more demanding qualifying conditions for admission seem to have had a greater impact on certain sub-groups among family immigrants. As this paper has demonstrated, this stratification is intertwined with the nationality and legal status of resident sponsors (cf. Morris 2003). Demanding admission policies, including requirements for permanent residence and adequate economic resources, are naturally harder to comply with for sponsors who are newly arrived immigrants with smaller means of self-support (Block 2015). These characteristics are indeed more common in the case of TCNs – especially those who have migrated from poor countries or fled from conflict or persecution – than in the case of nationals and other EU citizens. Accordingly, the legal status of resident sponsors, which is intertwined with the possession (or not) of financial and cultural capital, becomes increasingly important in determining opportunities for family reunification and family formation. As previous studies have suggested, moreover, such stratification can result in racialized patterns of immigrant selection (Bonjour & Duyvendak 2018; Schweitzer 2015).
Furthermore, my findings confirm the claim made in a growing body of literature that migration policy has become more selective across the board (de Haas, Natter & Vezzoli 2018; Helbling, Simon & Schmidt 2020). Still, as previous researchers have pointed out, there are good reasons to draw attention to family migration in particular, in view of recent reforms in migration policy (e.g. Bech, Borevi & Mouritsen 2017; Kofman 2018; Scholten et al. 2012; Wray et al. 2023). Given the overlapping and multi-layered nature of family related immigration, states have many policy options at their disposal to regulate and shape the admission and settlement of family immigrants. This becomes clear if we take into account the multidimensionality of admission policies, whereby varying requirements can be imposed both on resident sponsors and on incoming family members.
Finally, this paper has highlighted the need for analytical precision if we are to assess adequately how policies, whether by themselves or together with other factors, shape the size and composition of immigrant inflows. Since available data on migration policy now makes it possible to distinguish between policies that regulate different categories of entry (e.g. Helbling et al. 2017; Solano & Huddleston 2020), future research on comparative migration policies should devote greater attention to different categories of entry and devise different explanations for them. This relates, for example, to the interplay between different policy dimensions in connection with a particular immigrant category and to the ways in which policies can have varying effects on different sub-groups among immigrants. A welcome development in this regard would be data on immigrant inflows that distinguishes between different sub-groups within these broader categories (e.g. between cases of family reunification and of family formation). This would make it possible to analyse the differential impact of different policies in more detail.
Data Accessibility Statement
The data that support the findings of this study were derived from the following resources available in the public domain and listed in the references: Coppedge et al. (2021), Eurostat (2021a), Eurostat (2021b), ILO (2020), Nelson et al. (2020), Solano and Huddleston (2020). The policy datasets used (Coppedge et al. 2021; Nelson et al. 2020; Solano & Huddleston 2020) were accessed through the Demscore e-infrastructure (https://www.demscore.se/).
Additional File
The additional file for this article can be found as follows:
Appendices
Appendix
Table A1
Summary statistics, 2008-2019.
| VARIABLE | MEAN | STD. DEV. | MIN | MAX | N |
|---|---|---|---|---|---|
| Dependent variables | |||||
| Family migration (all) | 1.61 | 1.19 | 0.27 | 6.58 | 361 |
| Sponsor EU citizen | 0.56 | 0.53 | 0.003 | 3.52 | 354 |
| Sponsor non-EU citizen | 1.05 | 0.97 | 0.001 | 5.65 | 349 |
| Independent variables | |||||
| Policy variables | |||||
| Admission policies | 41.92 | 20. 14 | 0 | 100 | 355 |
| Residence-permit policies | 54.58 | 20.0 | 12.5 | 87.5 | 355 |
| Eligibility criteria | 60.95 | 29.38 | 0 | 100 | 355 |
| Qualifying conditions | 29.22 | 21.14 | 0 | 100 | 355 |
| Other independent variables | |||||
| Stocks of foreign-born | 0.12 | 0.08 | 0.01 | 0.47 | 351 |
| Immigrant inflow (excl. family) | 4.23 | 5.00 | 0.27 | 38.07 | 361 |
| Vote share of anti-immigration parties | 11.79 | 14.85 | 0 | 69.4 | 372 |
| Social-assistance rates (log) | 10710.6 | 5881.5 | 1701.2 | 31013.4 | 341 |
| Unemployment rate | 8.56 | 4.62 | 2.06 | 27.7 | 372 |
| GDP per capita (log) | 10.09 | 0.68 | 8.49 | 11.53 | 372 |
[i] Notes: Immigration rates and stocks of foreign-born persons concern third-country nationals (TCNs) and are measured per 1000 population. Policy variables are measured from 0–100, where higher values symbolize a higher degree of restrictiveness and lower values a lower one. Countries: Austria, Belgium, Bulgaria, Croatia, Cyprus, Czechia, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, the Netherlands, Norway, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden, Switzerland, and the United Kingdom.
Table A2
Fixed-effects regression models of family-migration rates in 29 European countries, 2008-2019.
| FAMILY MIGRATION (ALL) COUNTRY GROUP 1 | FAMILY MIGRATION (ALL) COUNTRY GROUP 2 | FAMILY MIGRATION (ALL) | FAMILY MIGRATION (ALL) | |
|---|---|---|---|---|
| MODEL 6 | MODEL 7 | MODEL 8 | MODEL 9 | |
| Admission policies t-1 | –0.0173** (0.00603) | 0.0100 (0.0164) | ||
| Eligibility criteria t-1 | –0.0143** (0.00583) | |||
| Qualifying conditions t-1 | –0.00614 (0.00537) | |||
| Residence-permit policies t-1 | 0.00599 (0.00751) | 0.00494 (0.00775) | 0.00181 (0.00724) | –0.00363 (0.00503) |
| Social-assistance rates t-1 | –3.65e–05 (6.30e–05) | 0.00017*** (4.03e–05) | 6.01e–05 (3.55e–05) | 6.75e–05** (3.28e–05) |
| Stocks of foreign-born t-1 | 9.707*** (2.411) | 5.819 (4.801) | 5.406 (3.233) | 5.451 (3.611) |
| Immigrant inflow (excl. family) | 0.147*** (0.0485) | 0.0255 (0.0223) | 0.0440** (0.0190) | 0.0413** (0.0181) |
| Unemployment rate t-1 | –0.0229 (0.0523) | –0.0481 (0.0453) | –0.0291 (0.0320) | –0.0240 (0.0346) |
| Vote share of anti-immigration parties | 0.00341 (0.00565) | –0.00616 (0.00614) | –0.00653 (0.00446) | –0.00167 (0.00399) |
| GDP per capita (log) t-1 | –0.413 (0.702) | –1.138 (1.366) | –0.736 (0.578) | –0.480 (0.592) |
| Constant | 5.478 | 9.756 | 8.312 | 5.235 |
| (7.084) | (13.36) | (5.888) | (5.995) | |
| Mean of dependent variable | 2.023 | 1.297 | 1.61 | 1.61 |
| Observations | 141 | 147 | 288 | 288 |
| R-squared | 0.555 | 0.381 | 0.349 | 0.331 |
| Number of countries | 14 | 15 | 29 | 29 |
[i] Notes: Immigration rates and stocks of foreign-born concern third-country nationals (TCNs) and are measured per 1000 population. The policy variables are measured from 0–100, where higher values symbolize a higher degree of restrictiveness and lower values a lower one. Country group 1 (Northern and Western European countries): Austria, Belgium, Denmark, Finland, France, Germany, Iceland, Ireland, Luxembourg, the Netherlands, Norway, Sweden, Switzerland, and the United Kingdom. Country group 2 (Central/Eastern and Southern European countries): Bulgaria, Croatia, Cyprus, Czechia, Estonia, Greece, Hungary, Italy, Latvia, Lithuania, Malta, Poland, Portugal, Romania, Slovakia, Slovenia, and Spain. ***p < 0.01; **p < 0.05; *p < 0.1.
Table A3
Robustness checks of family-migration rates in 29 European countries, 2008–2019.
| FAMILY MIGRATION (ALL) | SPONSOR EU CITIZEN | SPONSOR NON-EU CITIZEN | FAMILY MIGRATION (ALL) | IMMIGRANT INFLOW (EXCL. FAMILY) | ADMISSION POLICIES | |
|---|---|---|---|---|---|---|
| MODEL 10 | MODEL 11 | MODEL 12 | MODEL 13 | MODEL 14 | MODEL 15 | |
| Family migration (all) t-1 | 0.886*** (0.0469) | –0.951 (1.131) | ||||
| Sponsor EU citizen t-1 | 0.771*** (0.0539) | |||||
| Sponsor non-EU citizen t-1 | 0.890*** (0.0377) | |||||
| Admission policies t-1 | –0.0035*** (0.0009) | –0.0004 (0.0009) | –0.0033*** (0.0010) | –0.0608 (0.0382) | ||
| Admission policies t-2 | –0.01343* (0.00693) | |||||
| Residence-permit status t-1 | –0.0026* (0.00143) | –0.00061 (0.00072) | –0.00224* (0.00118) | –0.00029 (0.00437) | 0.0131 (0.0432) | |
| Social-assistance rates t-1 | 1.20e–05* (6.97e–06) | 6.04e–06 (4.83e–06) | 7.93e–06 (6.91e–06) | 0.00007** (0.00003) | –0.00015 (0.00035) | –0.00079 (0.00076) |
| Stocks of foreign-born t-1 | 1.037*** (0.373) | 0.913** (0.419) | 0.598** (0.264) | 4.96598 (4.56738) | 34.56 (43.91) | 58.94 (56.02) |
| Immigrant inflow (excl. family) | 0.0175*** (0.00674) | 0.00305 (0.00322) | 0.0135*** (0.00416) | 0.06169** (0.02646) | ||
| Unemployment rate t-1 | –0.00494 (0.00658) | 0.00194 (0.00437) | –0.00801 (0.00605) | –0.04239 (0.03885) | –0.0148 (0.139) | –0.326 (0.205) |
| GDP per capita (log) t-1 | –0.0376 (0.0805) | –0.0346 (0.0446) | –0.0238 (0.0762) | –1.02624* (0.59272) | 12.99* (6.925) | –11.42* (6.446) |
| Vote share of anti-immigration parties | –0.00163 (0.00125) | –0.00064 (0.00077) | –0.00092 (0.00118) | –0.00086 (0.00297) | ||
| Vote share of anti-immigration parties t-1 | 0.0701(0.066) | |||||
| Constant | 0.410 | 0.226 | 0.331 | 11.07794* | –127.8* | 155.9** |
| (0.786) | (0.431) | (0.748) | (6.06356) | (72.31) | (67.20) | |
| Fixed effects | No | No | No | Yes | Yes | Yes |
| Mean of dependent variable | 1.61 | 0.56 | 1.05 | 1.61 | 4.23 | 41.92 |
| Observations | 287 | 281 | 278 | 265 | 288 | 290 |
| R–squared | 0.574 | 0.272 | 0.485 | 0.3696 | 0.340 | 0.173 |
| Number of countries | 29 | 29 | 29 | 29 | 29 | 29 |
[i] Notes: Immigration rates and stocks of foreign-born persons concern third-country nationals (TCNs) and are measured per 1000 population. Policy variables are measured from 0–100, where higher values symbolize a higher degree of restrictiveness and lower values a lower one. ***p < 0.01; **p < 0.05; *p < 0.1.
Notes
[5] Eligibility criteria also include minimum age limits for sponsors and for family members. However, since this policy instrument is not included in MIPEX, it is not taken into account in this study.
[6] Note that the variable residence-permit policies concern incoming family migrants. Regulations connected to the legal status of resident sponsors, that is, whether they need to have a certain type of residence permit before their family members can immigrate – is captured by EC.
[7] The MIPEX index is used instead of data provided by Immigration Policies in Comparison (IMPIC) (Helbling et al. 2017), since the latter has available data only for the 1980 to 2010 period. Thus, using MIPEX makes it possible to combine policy data with detailed immigration data from Eurostat (2021a) for the 2008 to 2019 period.
[8] However, some policy instruments that states have at their disposal to regulate family migration are missing from MIPEX. These include minimum-age requirements, as well as demands that the resident sponsor have a certain type of residence permit (e.g., permanent residence) in order to apply for family reunification.
[12] The statistical models that include control variables exclude Croatia and Greece due to missing data on social-assistance rates for these countries (see Nelson et al. 2020).
[13] Austria, Belgium, Denmark, Finland, France, Germany, Iceland, Ireland, Luxembourg, the Netherlands, Norway, Sweden, Switzerland and the United Kingdom.
Acknowledgements
Previous versions of this study have been presented at the IMISCOE annual conference (July 7–9, 2021), the ECPR general conference (August 30–September 3, 2021), and the SWEPSA annual conference (September 29–October 1, 2021). I thank the participants in these seminars, as well as Axel Cronert, Joakim Palme, Karin Borevi, Per Strömblad, Kristoffer Jutvik, Frida Boräng and two anonymous reviewers, for providing helpful comments and advice.
Competing Interests
The author has no competing interests to declare.